Track your Data Science with Skore.

Evaluate, compare, and track your ML experiments. Built by the team that created and maintains scikit-learn. One line of code, comprehensive model evaluation, smart methodological guidance.

Sign up for Skore

Probabl maintains
scikit-learn.

Our team builds and maintains robust machine learning algorithms while keeping them simple and accessible, staying true to scikit-learn's founding philosophy of making predictive data analysis tools efficient and reusable in any context.

Beyond scikit-learn, we're expanding our impact across the entire data science pipeline: from where your data lives, with skrub handling the messy reality of heterogeneous tables and dataframes, to guiding you through the maze of experimentation with skore, helping data scientists move faster from raw data to validated, production-ready models.

This is some text inside of a div block.

Transparency

Understand how models work so you can trust and improve them.

This is some text inside of a div block.

Composabilité

Modular tools that fit your stack. No lock-in, ever.

This is some text inside of a div block.

Réutilisabilité

Past experiments become building blocks. Nothing is lost.

This is some text inside of a div block.

Science first

Start with the problem, not the tool. Methodology guides everything.

scikit-learn core team employed by Probabl
GV

Gael Varoquaux

Co-founder, CSO

OG

Olivier Grisel

Co-founder,
Core maintainer

GL

Guillaume Lemaitre

Co-founder,
Core maintainer

JdB

Jérémie du Boisberranger

Co-founder,
Core maintainer

AJ

Adrin Jalali

Co-founder,
Core maintainer

LE

Loïc Estève

Co-founder,
Core maintainer

FG

François Goupil

Co-founder,
Core contributor

AA

Arturo Amor

Co-founder,
Core contributor

SS

Stefanie Senger

Core maintainer

From an open-source initiative to the world's most used ML library

scikit-learn is created
2007

David Cournapeau publishes scikits.learn as a Google Summer of Code project.

First public release
2010

Gaël Varoquaux, Fabian Pedregosa, Alexandre Gramfort and Vincent Michel at Inria take leadership. Version 0.1 beta ships on February 1st. The project gets renamed to scikit-learn.

JMLR paper published
2011

The foundational paper is published in the Journal of Machine Learning Research. The community grows to hundreds ofcontributors. First coding sprint in Spain.

Consortium at the Inria Foundation
2018

A consortium of corporate sponsors is created to fund scikit-learn's long-term maintenance: Microsoft, BCG, AXA, BNPParibas Cardif, Intel, NVIDIA, Dataiku. Later joined by Chanel, Michelin, and Hugging Face among others. 42 milliondocumentation visits that year alone.

Version 1.0
2021

scikit-learn 1.0.0 ships after 2,100+ merged pull requests. A landmark release after 11 years of continuous development.

Probabl is founded
2023

The scikit-learn core team at Inria spins off to create Probabl, with support from the French government's France 2030program and Inria Participations.

Practitioner certifications and skolar
2024

Launch of the scikit-learn practitioner certifications and skolar, the reference learning platform for open-source machinelearning.

€13M seed, a European record for open source
2025

Probabl raises €13M, the largest seed round for an open-source company in Europe. scikit-learn 1.8 ships with nativeGPU support via the Array API.

skore launches, scikit-learn Central goes live
2026

Probabl officially ships skore, the Data Science platform by the scikit-learn founders. scikit-learn Central launches as theecosystem catalog and use case explorer. 4.1 billion cumulative downloads reached.

Tools and expertise from the source

We build products that encode the methodology and best practices our maintainers
have developed over 15 years. Every data scientist can benefit from that depth, from
day one.

The Data Science platform by the scikit-learn founders

Skore

Evaluate any scikit-learn compatible model in one line of code. skore automatically generates the right metrics, feature importance plots, and diagnostics for your use case, with smart methodological warnings that catch common pitfalls before they reach production.
 Automated evaluation reports
Cross-validation insights
Model comparison
Methodological guidance
Example 
import skore, skrub
from sklearn.linear_model import Ridge
model = skrub.tabular_pipeline(Ridge())
report = skore.evaluate(model, df, y, splitter=0.2)
report
import skore, skrub
from sklearn.linear_model import Ridge
model = skrub.tabular_pipeline(Ridge())
report = skore.evaluate(model, df, y, splitter=5)
report
import skore, skrub
from sklearn.linear_model import Ridge
from sklearn.ensemble import RandomForestRegressor
models = [
    skrub.tabular_pipeline(Ridge()),
    skrub.tabular_pipeline(RandomForestRegressor()),
]
comparison_report = skore.evaluate(models, df, y, splitter=5)
comparison_report
Sign up for Skore
The official scikit-learn training and certification

Skolar

Hands-on courses built by scikit-learn core developers, the same people who created the MOOC that reached 40,000+ learners worldwide. Validate your ML expertise with the only official scikit-learn certification.
New — Ecosystem Explorer
Scikit-learn Central
Discover packages, real-world use cases, and community rankingsacross the scikit-learn ecosystem.
Explore Catalog →

Use cases

This is some text inside of a div block.

Transparency

Understand how models work so you can trust and improve them.

This is some text inside of a div block.

Composabilité

Modular tools that fit your stack. No lock-in, ever.

This is some text inside of a div block.

Réutilisabilité

Past experiments become building blocks. Nothing is lost.

This is some text inside of a div block.

Science first

Start with the problem, not the tool. Methodology guides everything.

AI Agents

Credit Card Fraud Detection

Detect fraudulent transactions among millions of legitimate credit card purchases
banking
insurance
fraud detection
classification
scikit-learn
skore
AI Agents

Credit Card Fraud Detection

Detect fraudulent transactions among millions of legitimate credit card purchases
banking
insurance
fraud detection
classification
scikit-learn
skore

Stop shipping broken models. Start deciding together.

Track your first experiment in 5 minutes. No sign-up required, no vendor lock-in. Open source, built on scikit-learn.

Try Skore for Free